diff --git a/.github/workflows/strix.yml b/.github/workflows/strix.yml
index 9a76dbd57..6510febd3 100644
--- a/.github/workflows/strix.yml
+++ b/.github/workflows/strix.yml
@@ -855,7 +855,11 @@ jobs:
           VERTEX_LOCATION: ${{ secrets.VERTEX_LOCATION || 'us-central1' }}
           STRIX_TARGET_PATH: ${{ (github.event_name == 'pull_request_target' || github.event.client_payload.pr_number != '') && '__PR_SCOPE__' || './' }}
           STRIX_SOURCE_DIRS: ". backend frontend"
-          STRIX_REASONING_EFFORT: high
+          # OpenAI GPT-5.4's chat-completions endpoint rejects reasoning_effort
+          # when Strix supplies function tools. Keep high effort for providers
+          # that support it, but send the API-compatible neutral value to direct
+          # OpenAI scans so the security gate can complete.
+          STRIX_REASONING_EFFORT: ${{ steps.gate.outputs.provider_mode == 'openai_direct' && 'none' || 'high' }}
           STRIX_LLM_MAX_RETRIES: 1
           STRIX_TRANSIENT_RETRY_PER_MODEL: 2
           STRIX_TRANSIENT_RETRY_BACKOFF_SECONDS: 60
diff --git a/scripts/ci/strix_quick_gate.sh b/scripts/ci/strix_quick_gate.sh
index c9aa41545..7091b36cd 100755
--- a/scripts/ci/strix_quick_gate.sh
+++ b/scripts/ci/strix_quick_gate.sh
@@ -2551,6 +2551,7 @@ run_strix_once() {
 	local rc
 	local llm_api_base_value
 	local child_model
+	local child_reasoning_effort="${STRIX_REASONING_EFFORT:-high}"
 	local resolved_target_path
 	local timeout_seconds="$STRIX_PROCESS_TIMEOUT_SECONDS"
 	local total_budget_limited_timeout=0
@@ -2574,6 +2575,12 @@ run_strix_once() {
 		return 2
 	fi
 	child_model="$(child_model_for_api_base "$model" "$llm_api_base_value")"
+	if is_explicit_openai_model "$model"; then
+		# GPT-5.4 chat-completions rejects reasoning_effort when Strix
+		# supplies function tools; apply the override to every direct-OpenAI
+		# fallback, regardless of the selected primary provider.
+		child_reasoning_effort="none"
+	fi
 	if ! resolved_target_path="$(resolve_current_target_path "$TARGET_PATH")"; then
 		return 1
 	fi
@@ -2597,6 +2604,7 @@ run_strix_once() {
 	set -o pipefail
 	set +e
 	STRIX_CHILD_MODEL="$child_model" \
+	STRIX_CHILD_REASONING_EFFORT="$child_reasoning_effort" \
 	STRIX_CHILD_LLM_API_KEY="$child_llm_api_key" \
 	STRIX_CHILD_LLM_API_BASE="$llm_api_base_value" \
 	STRIX_CHILD_REPORTS_DIR="$ACTIVE_REPORTS_DIR" \
@@ -2671,7 +2679,6 @@ for key in (
 	"GEMINI_LOCATION",
 	"LLM_TIMEOUT",
     "STRIX_MEMORY_COMPRESSOR_TIMEOUT",
-    "STRIX_REASONING_EFFORT",
     "STRIX_LLM_MAX_RETRIES",
     "GOOGLE_CLOUD_PROJECT",
     "GCP_PROJECT",
@@ -2682,6 +2689,11 @@ for key in (
     value = os.environ.get(key)
     if value:
         child_env[key] = value
+child_reasoning_effort = os.environ.get("STRIX_CHILD_REASONING_EFFORT") or os.environ.get(
+    "STRIX_REASONING_EFFORT"
+)
+if child_reasoning_effort:
+    child_env["STRIX_REASONING_EFFORT"] = child_reasoning_effort
 llm_api_base = os.environ.get("STRIX_CHILD_LLM_API_BASE", "")
 if llm_api_base:
     child_env["LLM_API_BASE"] = llm_api_base
diff --git a/scripts/ci/test_strix_quick_gate.sh b/scripts/ci/test_strix_quick_gate.sh
index 3d3449dae..0d1318171 100755
--- a/scripts/ci/test_strix_quick_gate.sh
+++ b/scripts/ci/test_strix_quick_gate.sh
@@ -358,7 +358,7 @@ assert_strix_workflow_pr_trigger_hardened() {
 	assert_file_contains "$workflow_file" "NVIDIA_NIM_API_KEY is required for Strix NVIDIA NIM scans" "strix workflow fails closed when NVIDIA credentials are absent"
 	assert_file_contains "$workflow_file" 'PROVIDER_MODE: ${{ steps.gate.outputs.provider_mode }}' "strix workflow passes provider mode through env"
 	assert_file_not_contains "$workflow_file" '[ "${{ steps.gate.outputs.provider_mode }}" = "openai_direct" ]' "strix workflow does not interpolate provider mode inside shell condition"
-	assert_file_contains "$workflow_file" "STRIX_REASONING_EFFORT: high" "strix workflow uses high reasoning effort when the selected provider/model supports it"
+	assert_file_contains "$workflow_file" "STRIX_REASONING_EFFORT: \${{ steps.gate.outputs.provider_mode == 'openai_direct' && 'none' || 'high' }}" "strix workflow disables unsupported OpenAI chat-completions reasoning effort while retaining high effort for other providers"
 	assert_file_contains "$workflow_file" 'trimmed_openai_key="$(printf '"'"'%s'"'"' "$sanitized_openai_key" | sed '"'"'s/^[[:space:]]*//;s/[[:space:]]*$//'"'"')"' "strix workflow trims whitespace-only OpenAI keys before gate validation"
 	assert_file_contains "$workflow_file" 'printf '"'"'%s'"'"' "$trimmed" > "$llm_api_key_file"' "strix workflow writes trimmed provider API keys into the trusted input file"
 	assert_file_contains "$workflow_file" 'STRIX_LLM_DEFAULT_PROVIDER: ${{ steps.gate.outputs.provider_mode == '"'"'vertex_ai'"'"' && '"'"'vertex_ai'"'"' || steps.gate.outputs.provider_mode == '"'"'nvidia_nim'"'"' && '"'"'nvidia_nim'"'"' || '"'"'openai'"'"' }}' "strix workflow selects the correct default provider"
diff --git a/tests/test_opencode_agent_contract.py b/tests/test_opencode_agent_contract.py
index 1e3756169..b9aacd19e 100644
--- a/tests/test_opencode_agent_contract.py
+++ b/tests/test_opencode_agent_contract.py
@@ -1981,7 +1981,10 @@ def test_workflow_provisions_sandbox_tool_and_reviewer_agent():
     assert "falling back to current-head REST check-runs" in workflow
 
     strix_workflow = Path(".github/workflows/strix.yml").read_text(encoding="utf-8")
-    assert "STRIX_REASONING_EFFORT: high" in strix_workflow
+    assert (
+        "STRIX_REASONING_EFFORT: ${{ steps.gate.outputs.provider_mode == 'openai_direct' && 'none' || 'high' }}"
+        in strix_workflow
+    )
 
     prompt_template = Path("scripts/ci/opencode_review_prompt_template.md").read_text(
         encoding="utf-8"
@@ -2993,3 +2996,15 @@ def test_r_package_load_deferral_requires_current_head_r_cmd_check():
     assert (
         "if (!is.na(pkg) && !requireNamespace(pkg, quietly = TRUE))" not in workflow
     )
+
+
+def test_strix_direct_openai_fallback_overrides_child_reasoning_effort():
+    """Direct OpenAI fallbacks must not inherit an incompatible effort value."""
+    quick_gate = Path("scripts/ci/strix_quick_gate.sh").read_text(encoding="utf-8")
+
+    assert 'local child_reasoning_effort="${STRIX_REASONING_EFFORT:-high}"' in quick_gate
+    assert 'if is_explicit_openai_model "$model"; then' in quick_gate
+    assert 'child_reasoning_effort="none"' in quick_gate
+    assert 'STRIX_CHILD_REASONING_EFFORT="$child_reasoning_effort"' in quick_gate
+    assert 'child_reasoning_effort = os.environ.get("STRIX_CHILD_REASONING_EFFORT") or os.environ.get(' in quick_gate
+    assert 'child_env["STRIX_REASONING_EFFORT"] = child_reasoning_effort' in quick_gate